Value stream mapping, explained by drawing one
One machining plant: stamping, welding, assembly, shipping. As you scroll, the current-state map assembles itself the way a practitioner would draw it: customer first, data boxes, inventory triangles, the information flow that explains them, and the timeline that delivers the verdict. Then the same map is redesigned into a future state, and the timeline collapses.
A map of material and information
A value-stream map is not a flowchart. It draws two flows on one page, material along the bottom and information along the top, because the second one causes the first. The numbers are a worked example: a plant makes one bracket for one customer, and the numbers are round enough to check.
Start at the right edge — with demand.
A value-stream map is drawn customer-first. Say a plant ships a steering bracket to a vehicle plant: 18,400 pieces a month, two shifts, 460 available minutes each. Divide available time by demand and takt falls out: 55,200 s ÷ 920 pcs = 60 seconds. Every minute, the customer consumes one bracket. Everything else on this page answers to that number.
Walk the flow. Box every process.
Now walk the plant from shipping back upstream, and draw one box wherever material stops and starts again: Stamping → Welding → Assembly → Shipping. A box isn’t a department — it’s a segment of flow. If parts queue between two machines, those machines get separate boxes.
Fill the data boxes — with a stopwatch.
Under each box: cycle time, changeover, uptime, operators, and batch size — measured at the process, not pulled from the ERP. Notice stamping: 3 seconds a piece, but a 45-minute changeover, so it runs 5-day batches. That one number will explain most of this map.
Count what’s sitting between.
Wherever material waits, draw a triangle and count it. Then convert pieces to time at the demand rate: 4,600 stamped brackets ÷ 920 a day = 5 days of supply. Six days of coils, five of stampings, three of weldments, two of finished goods. Each triangle is the footprint of a problem you haven’t found yet.
Add the supplier and the trucks.
Coils arrive by truck weekly — which is why six days of steel sits by the door. Finished brackets leave daily. The map now shows every place material moves or waits, dock to dock.
Now the top half — and the plot twist.
Material is only half a value-stream map. Production control takes the customer’s forecast and sends a separate weekly schedule to every process. Each process obeys its own schedule — so each one produces whether or not the next one is ready. That’s a push system (the striped arrows), and it’s exactly where the triangles come from. Nobody chose the inventory; the information flow did.
Draw the ladder. Read the verdict.
Along the bottom, the timeline adds it up: waiting on the high rungs, working on the low ones. A bracket spends 16 days in the plant and 104 seconds being worked on — a process cycle efficiency of about 0.012%. That ratio is normal for a first map. The point isn’t shame; it’s that lead time lives in the triangles, not the boxes.
Find the overburden before you fix anything.
Compare every cycle time to takt (60 s). Stamping and welding fit easily. Assembly runs 62 s — over takt — so it falls behind all shift and overtime papers over the gap. Two different problems are now visible: assembly is overburdened, and stamping’s changeover forces batches. They need different countermeasures — and that’s what the future state is for.
The current state is a diagnosis. Now prescribe.
Sixteen days of lead time, 104 seconds of work, one overburdened process, and a changeover that forces batches. A future-state map answers four questions: what’s takt, where can material flow, where must it pull, and at which single point do you schedule?
Same plant, same 104 seconds of work, redesigned
The schedules come down, pull replaces push, and the timeline gets redrawn.
Schedule one point. Only one.
Pick the process that sets the rhythm for everything upstream — here, assembly — and send the schedule only there. The weekly schedules to stamping, welding, and shipping come off the wall. From now on, one point paces the stream; everyone else responds to signals, not spreadsheets.
Where flow can’t reach, put a supermarket.
Stamping shares its press with other parts, so it can’t flow one-to-one with welding. Instead: a small supermarket of stampings. Welding withdraws what it needs; each withdrawal sends a kanban telling stamping to replace exactly that. And because a supermarket is only as small as the changeover allows, an SMED workshop cuts C/O from 45 to 9 minutes — batches drop from 5 days to 1, and the supermarket holds 1.5 days instead of 5.
Between welding and assembly: FIFO.
These two can nearly flow, so no supermarket — a FIFO lane capped at 90 pieces. Sequence preserved, and when the lane is full, welding stops. The cap is the discipline. Meanwhile a kaizen burst marks the assembly fix: rebalance the work elements so assembly runs 58 s — under takt, with a little slack for variation.
The doors get the same treatment.
Shipping pulls from a one-day finished-goods supermarket — assembly builds only what shipping withdrew. Upstream, weekly coil deliveries become a daily milk run, and six days of steel becomes two. Every striped push arrow on the map is gone.
Redraw the ladder.
Lead time: 16 days → 4.6 days, a 71% cut. Every high rung on the ladder got shorter. Value-creating time barely moved (104 s → 100 s) — and that’s the real lesson of VSM: you didn’t make anyone work faster. You removed waiting. PCE is still a fraction of a percent, and that’s normal; the win is the 11 days of lead time — and the inventory, cash, and defect-feedback delay that went with them.
When the thing flowing is a patient or a claim
In offices and hospitals the map gains a third number: percent complete and accurate, how often a step receives work it can use without chasing anything down. %C&A multiplies down the chain the way uptime never does: five steps at 60–98% each can leave a system right-first-time only about 17% of the time. When handoffs of information dominate the stream, map %C&A alongside the two clocks. See the full map of an emergency department.
Now draw yours
The map you just watched took a pencil, a stopwatch and a walk. The free Kaizumi VSM Builder draws process boxes, triangles, kanban loops and the timeline for you, and there’s a PowerPoint template with the full symbol library if you’d rather map with your team on a wall.
Open the free VSM BuilderThe bracket plant, its four processes, and every figure on the map (18,400 pcs/month, takt 60 s, CT 3/39/62 s, 16 → 4.6 days) are a worked example built for teaching: round numbers you can recompute, not measured data from any real company. The method follows Rother and Shook, Learning to See (Lean Enterprise Institute, 1999).
Frequently asked
- What is value stream mapping?
- Value stream mapping (VSM) is a Lean method for drawing every step a product takes from supplier to customer on one page — material flow along the bottom, information flow along the top — with measured data at each step: cycle time, changeover, uptime, and inventory converted to days of supply. The map ends with a timeline ladder that totals lead time against value-creating time, which is the diagnosis a future-state redesign starts from.
- Why does a value-stream map show information flow, not just material flow?
- Because information flow causes material flow. When production control sends a separate schedule to every process, each process produces to its own schedule regardless of what the next one needs — that is push, and it is where the inventory triangles come from. Redesigning the stream means changing who gets scheduled (one pacemaker point) and letting kanban signals replace the other schedules, so the map must show both flows to explain either.
- How do you calculate takt time on a value-stream map?
- Divide available working time per day by customer demand per day. In the worked example: two shifts of 460 minutes give 55,200 seconds per day; demand of 18,400 pieces a month over 20 working days is 920 pieces a day; 55,200 ÷ 920 = a takt of 60 seconds. Every process cycle time on the map is then read against that number — a process over takt (like assembly at 62 s) is overburdened and falls behind every shift.
- How do you calculate lead time and Process Cycle Efficiency from the timeline ladder?
- Convert each inventory count to days of supply by dividing by daily demand (4,600 stamped pieces ÷ 920 a day = 5 days), then sum the days on the high rungs of the ladder for lead time, and sum the processing times on the low rungs for value-creating time. In the example: 6 + 5 + 3 + 2 = 16 days of lead time against 104 seconds of work — a Process Cycle Efficiency of roughly 0.012%. Ratios that small are normal for a first current-state map; the lesson is that lead time lives in the inventory, not in the processing.
- What is the difference between a supermarket and a FIFO lane on a value-stream map?
- Both are pull mechanisms with a cap, used where continuous flow is not possible. A supermarket is a small store of every part the downstream process might withdraw; each withdrawal releases a kanban telling the upstream process to replace exactly what was taken — used where a process runs batches or serves several streams (like a shared stamping press). A FIFO lane is a first-in-first-out chute between two processes that nearly flow: sequence is preserved, the quantity is capped, and when the lane is full the upstream process stops.
- What is the pacemaker process in a future-state map?
- The single point in the stream that receives the production schedule. Everything downstream of the pacemaker flows or moves FIFO; everything upstream is replenished by pull signals from supermarkets. Scheduling one point instead of every process is what removes the push arrows — in the worked example the pacemaker is assembly, and the weekly schedules to stamping, welding, and shipping come off the wall.